Research map / Graph learning / Graph self-supervised learning
GALOPA: Graph Transport Learning with Optimal Plan Alignment
Yejiang Wang, Yuhai Zhao, Daniel Zhengkui Wang, Ling Li
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Align optimal transport plan instead of OT distance (ablation) | GALOPA · CORA | accuracy: 0.795 -> 0.842 | Improved |
| Align optimal transport plan instead of OT distance (ablation) | GALOPA · CITESEER | accuracy: 0.689 -> 0.743 | Improved |
| Align optimal transport plan instead of OT distance (ablation) | GALOPA · PROTEINS | accuracy: 0.733 -> 0.769 | Improved |
| Align optimal transport plan instead of OT distance (ablation) | GALOPA · MUTAG | accuracy: 0.851 -> 0.911 | Improved |
| Use node attributes only (ignore edge structure, rho=0, sigma=1) in the graph-space calibration signal (ablation) | GALOPA · CORA | accuracy: 0.842 -> 0.776 | Hurt |
| Use node attributes only in the graph-space calibration signal (rho=0, sigma=1) (ablation) | GALOPA · CITESEER | accuracy: 0.743 -> 0.647 | Hurt |
| Vary sigma from 1 to 0.5 to 0 (node attribute vs edge weighting) with rho!=0 (ablation) | GALOPA · CORA | accuracy: 0.842 -> 0.835 -> 0.823 | No clear effect |
| Vary sigma from 1 to 0.5 to 0 with rho!=0 (ablation) | GALOPA · MUTAG | accuracy: 0.91 -> 0.911 -> 0.868 | Mixed |
| Remove 80% of edges and destroy 80% of node attributes in augmentation (ablation) | GALOPA · CITESEER | accuracy: 0.743 -> 0.742 | No clear effect |
| Heavy augmentation perturbation (edge perturbation plus feature masking) (ablation) | GALOPA · PROTEINS | accuracy | No clear effect |
| Use GALOPA instead of best unsupervised baseline GCA/SimGRACE (ablation) | GALOPA vs baselines · PUBMED | accuracy (%): 80.89 -> 84.57 | Improved |
| Use GALOPA instead of SimGRACE (ablation) | GALOPA vs baselines · TU graph datasets average | accuracy (%): 77.31 -> 78.94 | Improved |
Closest papers by meaning
- Self-Supervised Node Representation Learning via Node-to-Neighbourhood Alignment (TPAMI 2024 · similarity 0.89)
- Spectral Augmentation for Self-Supervised Learning on Graphs (ICLR 2023 · similarity 0.89)
- Gromov-Wasserstein Learning for Graph Matching and Node Embedding (ICML 2019 · similarity 0.88)
- HGOT: Self-supervised Heterogeneous Graph Neural Network with Optimal Transport (ICML 2025 · similarity 0.88)
- Hypergraph Learning for Unsupervised Graph Alignment via Optimal Transport (AAAI 2025 · similarity 0.88)
- Exploitation of a Latent Mechanism in Graph Contrastive Learning: Representation Scattering (NeurIPS 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
